1. Field of the Invention
The present invention relates to an image processing apparatus, an image processing method, and a computer program product.
2. Description of the Related Art
In image forming apparatuses, the needs for generating images with less roughness and low graininess in image quality has been increased. To meet such needs, an image processing method has been devised that uses an error diffusion process of periodically amplifying a threshold for quantizing an input gray-scale value, which is input image data, in an image space (refer to Japanese Patent Application Laid-Open No. 2001-128004).
Also, an image processing scheme has been suggested in which, when the density level of a target pixel for image processing is equal to or lower than an upper-limit density region, as the density level of the target pixel is higher, a region in which a print dot can occur is expanded (refer to Japanese Patent Application Laid-Open No. 2003-153001). According to this technology, an unbalanced local dot occurrence in a low-density region is suppressed, thereby suppressing the occurrence of texture.
However, in the image processing method disclosed in Japanese Patent Application Laid-Open No. 2001-128004, although a high-density image with excellent graininess can be generated compared with a simple error diffusion method, an image with excellent graininess cannot be obtained in a low-density image region.
FIG. 23 is an example of a dither threshold table according to a conventional example. For example, when binary quantization is performed on data of 8 bits (0 to 255) of a low-density image by using the dither threshold table shown in FIG. 23, both states of dot occurrence and non-occurrence generally occur only at pixel positions to which the lowest threshold of 72 is applied, and the probability of occurrence of the state of dot occurrence is low at other positions. That is, an output image is generated in an image space with a pixel density with the threshold of 72, which means that an output image is represented in a low-density image space compared with the case of, as in a general error diffusion process, using the same threshold at all pixel positions. This is thought to decrease dispersibility and increase graininess.
On the other hand, when binary quantization is performed on a high-density image region, both states of dot occurrence and non-occurrence generally occur only at pixel positions to which the highest threshold of 192 is applied, and the probability of occurrence of the state of dot non-occurrence is low at other positions. That is, an output image is generated in an image space with the density of the pixels with the threshold of 192. This decreases dispersibility and increases graininess.
Moreover, in the image processing scheme disclosed in Japanese Patent Application Laid-Open No. 2003-153001, in image regions other than those with low density, an output image identical to that obtained in a general error diffusion method is obtained. Therefore, as suggested in Japanese Patent Application Laid-Open No. 2001-128004, graininess and stability are inferior compared with those achieved by a mainly-dither process.